Such is the demand for Nvidia’s graphics processing units to advance artificial intelligence that Larry Ellison, the billionaire co-founder of Oracle, said he had “begged” for more.
Jensen Huang, Nvidia’s chief executive, was taken to dinner at Nobu Palo Alto by Ellison and Elon Musk, the world’s richest man, in September.
“I would describe the dinner as Oracle, me and Elon begging Jensen for GPUs,” Ellison later told analysts. “Please take our money. Please take our money. By the way, I got dinner. No, no, take more of it. We need you to take more of our money please.”
Jensen Huang, Nvidia’s co-founder and chief executive
PATRICK T FALLON/AFP/GETTY IMAGES
Nvidia has positioned itself as the most important supplier in the artificial intelligence revolution. The maker of graphics processing units, first developed for video gaming, has recorded extraordinary results since the public launch of ChatGPT in late 2022, which broadened the use case for its chips and triggered a wave of investment in AI infrastructure.
Demand for Nvidia’s chips leaped, with sales led by its H100 chip design, which can be used to run the large-language models that power generative AI programs such as ChatGPT.
The technology propelled Nvidia’s valuation above $1 trillion for the first time in June 2023. Last year, it raced past the $2 trillion and $3 trillion milestones in February and June, respectively.
On Wednesday its valuation surpassed $4 trillion in intraday trading, the first time any public company had done so.
What happened to Deepseek fears?
Investors have shaken off worries from January when Deepseek, a Chinese artificial intelligence start-up, said it was even more efficient than OpenAI, the owner of ChatGPT, and had tech that was just as powerful but ran on fewer chips.
Huang has dismissed concerns about chip demand, arguing that, instead of slowing down, demand for computing power to handle reasoning and agentic AI applications — which have more decision-making power and autonomy — will be higher than anticipated.
• Why rivals can’t compete with Nvidia’s chips for everything
“The amount of computation we need at this point as a result of agentic AI, as a result of reasoning, is easily a hundred times more than we thought we needed this time last year,” he said in March.
Nvidia technology has been vital to AI development
MUSTAFA YALCIN/ANADOLU/GETTY IMAGES
Big tech companies including Microsoft, Amazon, Alphabet and Meta Platforms have remained committed to investing billions of dollars to build expensive AI networks.
This week, Bernstein research estimated that total investment in planned new data centres driven by AI demand had risen to $750 billion — a positive signal for Nvidia, which offers the full computing “stack” required for AI, including hardware systems and the software needed to optimise their performance.
Nvidia continues to innovate with more powerful chips that can perform the complex tasks required by generative AI platforms. The Hopper H100 is being replaced by the Blackwell range, which allows two and a half times Hopper’s performance in training AI, according to Nvidia.
In the first quarter to April 27, Nvidia reported revenue of $44.06 billion, up 69 per cent on the year and beating estimates of $43.3 billion. Net income rose 26 per cent to $18.8 billion, beating estimates of $18.2 billion.
Analysts expect Nvidia to report sales of $45.6 billion for the second quarter, up 52 per cent on the same period last year. Net income is forecast to rise 47 per cent to $24.4 billion.
• Jensen Huang: the rise and rage of Nvidia’s visionary billionaire
What about China?
Nvidia was forced to write off inventory earmarked for China in April when the Trump administration put export limits on its H20 chip, designed specifically for the Chinese market. The company said the new restrictions would result in $5.5 billion in charges. The charges meant that Nvidia’s profit margin fell from 73 per cent to a still impressive 60.5 per cent in the last quarter.
Bullish analysts believe Nvidia can continue to grow by expanding in other markets.
The market for data centres is expected to continue expanding
JASON ALDEN/BLOOMBERG/GETTY IMAGES
Atif Malik, an analyst at Citi, raised his price target for Nvidia stock to $190 from $180 this week, citing higher demand from sovereign countries for AI, as governments invest heavily in the technology. Malik expects sovereign AI demand to drive the total addressable market for data-centre semiconductor AI to more than half a trillion dollars by 2028.
Is it a bubble?
Nvidia’s earnings multiple has already deflated from two years ago, when it peaked at 209 in July 2023. It has since fallen to 53.
At some point demand for GPUs will start to become saturated. However, semiconductor experts believe there will be several years of more growth ahead.
Jon Peddie, a former Nvidia adviser, said: “There’ll be a levelling off once enough people get enough GPUs, but I think that won’t be for quite a while. Our guesstimate is that we’ve still got two to three years of ridiculous growth ahead of us, and then it’s going to roll off, but roll off slowly.”
Investors should expect more volatility driven by geopolitics, tariffs and product transitions. However, over time the chip designer’s sales and shares should continue to rise.
Advice: Hold
Why: Nvidia’s growth story still has room to run.